Setting_up

Setting up grids

There are two methods for setting up grids to run parallel executions in Phoenix. The Compute Grid page of the Preferences dialog is one method (see the “Compute Grid preferences” section). The other is editing the configuration files directly, either as an administrator or user (see the “Configuration files” section).

Phoenix supports the following modes of parallelization:

Parallel execution

MultiCore (Windows, RHEL 8, Ubuntu 22): Parallelize (single models) across multiple cores on a single machine (i.e., 32-core RHEL 8, Ubuntu 22, or Windows computer).

MPI (Windows, RHEL 8, Ubuntu 22): Parallelize (single models and/or single model parallelized by subject) across multiple cores on an MPI Cluster.

Parallel execution of single model runs (e.g., bootstrap and shotgun covariate search) on a grid

LSF (Windows, RHEL 8, Ubuntu 22): Individual models are not parallelized across multiple cores.

SGE (RHEL 8, Ubuntu 22): Individual models are not parallelized across multiple cores.

TORQUE (RHEL 8, Ubuntu 22)

SLURM (RHEL 8, Ubuntu 22)

Parallel execution and parallelization across multiple cores of all model runs on a grid.

All models are run simultaneously and are parallelized across multiple cores. Phoenix determines the optimal parallelization strategy for each model run. Supports all Phoenix run modes.

LSF_MPI (RHEL 8, Ubuntu 22)

SGE_MPI (RHEL 8, Ubuntu 22)

TORQUE_MPI (RHEL 8, Ubuntu 22)

SLURM_MPI (RHEL 8, Ubuntu 22)

Example grid definitions for different parallel modes can be found in the PhoenixParallelExecutionSettings.txt file and modified with company-specific settings. (Refer to the “Configuration files” section for details on modifying this file.)

Note: When a grid is selected for executing an NLME object, loading the grid can take some time and it may seem that the application has stopped working.

Make sure that there is adequate disk space on the grid for execution of all jobs. A job will fail on the grid if it runs out of disk space.